On 21.06.23 04:58, David Hendricks wrote:
On Tue, Jun 20, 2023 at 4:41 PM Peter Stuge peter@stuge.se wrote:
ron minnich wrote:
And, yes, no question, this is an activity that likely occurs less than it should. Such is our industry.
Such is project policy. Maybe because it's the lowest common denominator in industry.
The project's policy is to remove code that becomes difficult to maintain, whether due to bitrot or technical barriers to refactoring.
I don't see this "becomes difficult to maintain" much. Much of what we removed in the last years had either an unfortunate design (cf. FSP 1.0) and or was already merged in a difficult to maintain state (cf. AGESA). IMO, this is where we should consider to learn from, not just make it policy and go on.
It is not possible to know, a priori, what those common pieces will be.
I think this is where we fundamentally disagree. I think common pieces and their interfaces can be recognized based on the hard IP blocks and their interfaces plus some creative thinking based on development experience.
I'm pretty sure the Intel engineers who wrote the Archer City CRB code didn't have IBM's and ByteDance's server specs handy or know how they would use the code.
AFAICT, the less board-specific code in archercity_crb/ is platform specific. This may be an important point: FWIW, we focus more on review for the platform code, and are more lax about the board ports. Maybe we should try to keep an eye on it, so that platform code doesn't sneak into board ports.
A bit OT, and sorry, if this is an odd question: Wasn't this the point of OCP, that they can talk to each other? If it's possible, we could try to talk earlier about the code, before it's written and pushed upstream later.
Nico